motion estimation for video coding

[new in imaging]

Hi,

in motion estimation for video coding the commonly used test sequences are foreman, football, miss america, claire etc.

While working on ME algorithms it is commonly seen that the same sequences are used first for analysis, and then based on that analysis the algorithm is devised/

then finally the same sequences are used for testing the algorithm

Do we need to use different sequences for testing or analysis, or its fine to use same for both purposes.

Thanks

[Dark Shikari]

in motion estimation for video coding the commonly used test sequences are foreman, football, miss america, claire etc.Which is unfortunate, because such low-resolution sequences are totally useless for any serious motion estimation algorithms. This probably explains why only 4 or 5 particularly useful motion estimation algorithms have ever been created.Do we need to use different sequences for testing or analysis, or its fine to use same for both purposes.Ideally, you want to use different sequences. Also, you should always use at least 720p or higher, because using CIF/similar sequences for motion estimation tests is a recipe for disaster (also known as the "Novel ______-shaped motion estimation for H.264/AVC video coding"-style of motion estimation algorithm).